I talked to Jim and he and I had exactly the same comments.  We love
this minus the numbers you pull out of an orifice ;-)

To be exact:
> -     sc->sc_reply_post_qdepth = 128;
> -     sc->sc_request_depth = 128;
> -     sc->sc_num_reply_frames = 63;
> -     sc->sc_reply_free_qdepth = 64;
> +     sc->sc_reply_post_qdepth = 8192;
> +     sc->sc_request_depth = 1024;
> +     sc->sc_num_reply_frames = 1151;
> +     sc->sc_reply_free_qdepth = 1152;

Both choices are bad.  This needs to come from the firmware and not made
up as we do now.  Jim's values are conservative, yours are probably ok
for *your* chip but no telling about others.  I think the port facts
pages has these.

I can go ahead and commit the rest.  Agree?

On Mon, Feb 08, 2010 at 07:13:36PM +0300, Mike Belopuhov wrote:
> Good day,
> 
> the following diff...
> 
> - implements bioctl support;
> - fixes hot-un-plugging w/ softeps;
> - improves performance;
> - fixes IPL levels;
> - fixes lots of small things;
> - does a little bit of cleanup;
> - fixes NOWAIT/WAITOK;
> - disables useless/unused Driver Persistent Mapping code that prevents
>   driver to work properly on our board;
> - maybe something else.
> 
> I thought about splitting it up in several pieces, but then I thought:
> "What the heck!  Lets beat the devil out of it!"
> 
> Enjoy and be merry.
> 
> Index: mpii.c
> ===================================================================
> RCS file: /mount/cvsdev/cvs/openbsd/src/sys/dev/pci/mpii.c,v
> retrieving revision 1.5
> diff -u -p -u -p -r1.5 mpii.c
> --- mpii.c    1 Dec 2009 00:09:03 -0000       1.5
> +++ mpii.c    8 Feb 2010 15:51:40 -0000
> @@ -30,6 +30,7 @@
>  #include <sys/kernel.h>
>  #include <sys/rwlock.h>
>  #include <sys/sensors.h>
> +#include <sys/tree.h>
>  
>  #include <machine/bus.h>
>  
> @@ -1019,7 +1020,6 @@ struct mpii_cfg_hdr {
>  #define MPII_CONFIG_REQ_PAGE_TYPE_MANUFACTURING              (0x09)
>  #define MPII_CONFIG_REQ_PAGE_TYPE_RAID_PD            (0x0a)
>  #define MPII_CONFIG_REQ_PAGE_TYPE_EXTENDED           (0x0f)
> -#define MPII_CONFIG_REQ_PAGE_TYPE_DRIVER_MAPPING     (0x17)
>  } __packed;
>  
>  struct mpii_ecfg_hdr {
> @@ -1030,6 +1030,9 @@ struct mpii_ecfg_hdr {
>  
>       u_int16_t               ext_page_length;
>       u_int8_t                ext_page_type;
> +#define MPII_CONFIG_REQ_PAGE_TYPE_SAS_DEVICE         (0x12)
> +#define MPII_CONFIG_REQ_PAGE_TYPE_RAID_CONFIG                (0x16)
> +#define MPII_CONFIG_REQ_PAGE_TYPE_DRIVER_MAPPING     (0x17)
>       u_int8_t                reserved2;
>  } __packed;
>  
> @@ -1269,52 +1272,45 @@ struct mpii_cfg_fc_device_pg0 {
>       u_int8_t                current_bus;
>  } __packed;
>  
> +#define MPII_CFG_RAID_VOL_ADDR_HANDLE                (1<<28)
> +
>  struct mpii_cfg_raid_vol_pg0 {
>       struct mpii_cfg_hdr     config_header;
>  
> -     u_int8_t                volume_id;
> -     u_int8_t                volume_bus;
> -     u_int8_t                volume_ioc;
> +     u_int16_t               volume_handle;
> +     u_int8_t                volume_state;
> +#define MPII_CFG_RAID_VOL_0_STATE_MISSING            (0x00)
> +#define MPII_CFG_RAID_VOL_0_STATE_FAILED             (0x01)
> +#define MPII_CFG_RAID_VOL_0_STATE_INITIALIZING               (0x02)
> +#define MPII_CFG_RAID_VOL_0_STATE_ONLINE             (0x03)
> +#define MPII_CFG_RAID_VOL_0_STATE_DEGRADED           (0x04)
> +#define MPII_CFG_RAID_VOL_0_STATE_OPTIMAL            (0x05)
>       u_int8_t                volume_type;
> +#define MPII_CFG_RAID_VOL_0_TYPE_RAID0                       (0x00)
> +#define MPII_CFG_RAID_VOL_0_TYPE_RAID1E                      (0x01)
> +#define MPII_CFG_RAID_VOL_0_TYPE_RAID1                       (0x02)
> +#define MPII_CFG_RAID_VOL_0_TYPE_RAID10                      (0x05)
> +#define MPII_CFG_RAID_VOL_0_TYPE_UNKNOWN             (0xff)
>  
> -     u_int8_t                volume_status;
> -#define MPII_CFG_RAID_VOL_0_STATUS_ENABLED           (1<<0)
> -#define MPII_CFG_RAID_VOL_0_STATUS_QUIESCED          (1<<1)
> -#define MPII_CFG_RAID_VOL_0_STATUS_RESYNCING         (1<<2)
> -#define MPII_CFG_RAID_VOL_0_STATUS_ACTIVE            (1<<3)
> -#define MPII_CFG_RAID_VOL_0_STATUS_BADBLOCK_FULL     (1<<4)
> -     u_int8_t                volume_state;
> -#define MPII_CFG_RAID_VOL_0_STATE_OPTIMAL            (0x00)
> -#define MPII_CFG_RAID_VOL_0_STATE_DEGRADED           (0x01)
> -#define MPII_CFG_RAID_VOL_0_STATE_FAILED             (0x02)
> -#define MPII_CFG_RAID_VOL_0_STATE_MISSING            (0x03)
> -     u_int16_t               reserved1;
> +     u_int32_t               volume_status;
>  
>       u_int16_t               volume_settings;
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_WRITE_CACHE_EN  (1<<0)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_OFFLINE_SMART_ERR       (1<<1)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_OFFLINE_SMART   (1<<2)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_AUTO_SWAP               (1<<3)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_HI_PRI_RESYNC   (1<<4)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_PROD_SUFFIX     (1<<5)
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_FAST_SCRUB              (1<<6) /* 
> obsolete */
> -#define MPII_CFG_RAID_VOL_0_SETTINGS_DEFAULTS                (1<<15)
>       u_int8_t                hot_spare_pool;
> -     u_int8_t                reserved2;
> -
> -     u_int32_t               max_lba;
> +     u_int8_t                reserved1;
>  
> -     u_int32_t               reserved3;
> +     u_int64_t               max_lba;
>  
>       u_int32_t               stripe_size;
>  
> -     u_int32_t               reserved4;
> +     u_int16_t               block_size;
> +     u_int16_t               reserved2;
>  
> -     u_int32_t               reserved5;
> +     u_int8_t                phys_disk_types;
> +     u_int8_t                resync_rate;
> +     u_int16_t               data_scrub_rate;
>  
>       u_int8_t                num_phys_disks;
> -     u_int8_t                data_scrub_rate;
> -     u_int8_t                resync_rate;
> +     u_int16_t               reserved3;
>       u_int8_t                inactive_status;
>  #define MPII_CFG_RAID_VOL_0_INACTIVE_UNKNOWN         (0x00)
>  #define MPII_CFG_RAID_VOL_0_INACTIVE_STALE_META              (0x01)
> @@ -1327,9 +1323,10 @@ struct mpii_cfg_raid_vol_pg0 {
>  } __packed;
>  
>  struct mpii_cfg_raid_vol_pg0_physdisk {
> -     u_int16_t               reserved;
> +     u_int8_t                raid_set_num;
>       u_int8_t                phys_disk_map;
>       u_int8_t                phys_disk_num;
> +     u_int8_t                reserved;
>  } __packed;
>  
>  struct mpii_cfg_raid_vol_pg1 {
> @@ -1351,6 +1348,8 @@ struct mpii_cfg_raid_vol_pg1 {
>       u_int32_t               reserved3;
>  } __packed;
>  
> +#define MPII_CFG_RAID_PHYS_DISK_ADDR_NUMBER          (1<<28)
> +
>  struct mpii_cfg_raid_physdisk_pg0 {
>       struct mpii_cfg_hdr     config_header;
>  
> @@ -1369,45 +1368,50 @@ struct mpii_cfg_raid_physdisk_pg0 {
>  
>       u_int32_t               reserved1;
>  
> -     u_int8_t                ext_disk_id[8];
> -
> -     u_int8_t                disk_id[16];
> -
>       u_int8_t                vendor_id[8];
>  
>       u_int8_t                product_id[16];
>  
>       u_int8_t                product_rev[4];
>  
> -     u_int8_t                info[32];
> +     u_int8_t                serial[32];
> +
> +     u_int32_t               reserved2;
> +
> +     u_int8_t                phys_disk_state;
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_NOTCONFIGURED  (0x00)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_NOTCOMPATIBLE  (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_OFFLINE                (0x02)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_ONLINE         (0x03)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_HOTSPARE               (0x04)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_DEGRADED               (0x05)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_REBUILDING     (0x06)
> +#define MPII_CFG_RAID_PHYDISK_0_STATE_OPTIMAL                (0x07)
> +     u_int8_t                offline_reason;
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_MISSING              (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_FAILED               (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_INITIALIZING (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_REQUESTED    (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_FAILEDREQ    (0x01)
> +#define MPII_CFG_RAID_PHYDISK_0_OFFLINE_OTHER                (0xff)
>  
> -     u_int8_t                phys_disk_status;
> +     u_int8_t                incompat_reason;
> +     u_int8_t                phys_disk_attrs;
> +
> +     u_int32_t               phys_disk_status;
>  #define MPII_CFG_RAID_PHYDISK_0_STATUS_OUTOFSYNC     (1<<0)
>  #define MPII_CFG_RAID_PHYDISK_0_STATUS_QUIESCED              (1<<1)
> -     u_int8_t                phys_disk_state;
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_ONLINE         (0x00)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_MISSING                (0x01)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_INCOMPAT               (0x02)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_FAILED         (0x03)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_INIT           (0x04)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_OFFLINE                (0x05)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_HOSTFAIL               (0x06)
> -#define MPII_CFG_RAID_PHYDISK_0_STATE_OTHER          (0xff)
> -     u_int16_t               reserved2;
>  
> -     u_int32_t               max_lba;
> +     u_int64_t               dev_max_lba;
> +
> +     u_int64_t               host_max_lba;
> +
> +     u_int64_t               coerced_max_lba;
>  
> -     u_int8_t                error_cdb_byte;
> -     u_int8_t                error_sense_key;
> +     u_int16_t               block_size;
>       u_int16_t               reserved3;
>  
> -     u_int16_t               error_count;
> -     u_int8_t                error_asc;
> -     u_int8_t                error_ascq;
> -
> -     u_int16_t               smart_count;
> -     u_int8_t                smart_asc;
> -     u_int8_t                smart_ascq;
> +     u_int32_t               reserved4;
>  } __packed;
>  
>  struct mpii_cfg_raid_physdisk_pg1 {
> @@ -1489,7 +1493,16 @@ struct mpii_cfg_sas_dev_pg0 {
>  #define MPII_CFG_SAS_DEV_0_FLAGS_UNSUPPORTED         (1<<8)
>  #define MPII_CFG_SAS_DEV_0_FLAGS_SATA_SETTINGS               (1<<9)
>       u_int8_t                physical_port;
> -     u_int8_t                reserved;
> +     u_int8_t                max_port_conn;
> +
> +     u_int64_t               device_name;
> +
> +     u_int8_t                port_groups;
> +     u_int8_t                dma_group;
> +     u_int8_t                ctrl_group;
> +     u_int8_t                reserved1;
> +
> +     u_int64_t               reserved2;
>  } __packed;
>  
>  struct mpii_cfg_bios_pg2 {
> @@ -1518,6 +1531,8 @@ struct mpii_cfg_bios_pg2 {
>       u_int32_t               c;
>  } __packed;
>  
> +#define MPII_CFG_RAID_CONFIG_ACTIVE_CONFIG           (2<<28)
> +
>  struct mpii_cfg_raid_config_pg0 {
>       struct  mpii_ecfg_hdr   config_header;
>  
> @@ -1871,6 +1886,7 @@ struct mpii_softc {
>       u_int16_t               sc_vd_id_low;
>       u_int16_t               sc_vd_id_hi;
>       u_int16_t               sc_pd_id_start;
> +     u_int16_t               sc_pd_count;
>       u_int16_t               sc_vd_count;
>       u_int8_t                sc_num_channels;
>       /* XXX not sure these below will stay */
> @@ -1949,7 +1965,8 @@ static const struct pci_matchid mpii_dev
>  int
>  mpii_pci_match(struct device *parent, void *match, void *aux)
>  {
> -     return (pci_matchbyid(aux, mpii_devices, nitems(mpii_devices)));
> +     return (pci_matchbyid(aux, mpii_devices,
> +         sizeof(mpii_devices) / sizeof(mpii_devices[0])));
>  }
>  
>  void
> @@ -2076,7 +2093,7 @@ struct scsi_adapter mpii_switch = {
>       mpii_minphys, 
>       mpii_scsi_probe, /* XXX JPG scsi_probe may prove useful for mapping 
> nonsense */
>       NULL,
> -     NULL
> +     mpii_scsi_ioctl
>  };
>  
>  struct scsi_device mpii_dev = {
> @@ -2101,7 +2118,8 @@ void            mpii_push_replies(struct mpii_soft
>  void         mpii_start(struct mpii_softc *, struct mpii_ccb *);
>  int          mpii_complete(struct mpii_softc *, struct mpii_ccb *, int);
>  int          mpii_poll(struct mpii_softc *, struct mpii_ccb *, int);
> -int          mpii_reply(struct mpii_softc *);
> +int          mpii_reply(struct mpii_softc *,
> +                 struct mpii_reply_descriptor *);
>  
>  void         mpii_init_queues(struct mpii_softc *);
>  
> @@ -2169,13 +2187,16 @@ void          mpii_reorder_vds(struct mpii_softc
>  void         mpii_reorder_boot_device(struct mpii_softc *);
>  
>  #if NBIO > 0
> -int          mpii_bio_get_pg0_raid(struct mpii_softc *, int);
>  int          mpii_ioctl(struct device *, u_long, caddr_t);
>  int          mpii_ioctl_inq(struct mpii_softc *, struct bioc_inq *);
>  int          mpii_ioctl_vol(struct mpii_softc *, struct bioc_vol *);
>  int          mpii_ioctl_disk(struct mpii_softc *, struct bioc_disk *);
> -int          mpii_ioctl_setstate(struct mpii_softc *, struct bioc_setstate 
> *);
> +int          mpii_bio_hs(struct mpii_softc *, struct bioc_disk *, int,
> +                 u_int8_t, int *);
> +int          mpii_bio_disk(struct mpii_softc *, struct bioc_disk *,
> +                 u_int8_t);
>  #ifndef SMALL_KERNEL
> + int         mpii_bio_volstate(struct mpii_softc *, struct bioc_vol *);
>  int          mpii_create_sensors(struct mpii_softc *);
>  void         mpii_refresh_sensors(void *);
>  #endif /* SMALL_KERNEL */
> @@ -2193,11 +2214,11 @@ void          mpii_refresh_sensors(void *);
>  #define mpii_reply_waiting(s)        ((mpii_read_intr((s)) & 
> MPII_INTR_STATUS_REPLY)\
>                                   == MPII_INTR_STATUS_REPLY)
>  
> -#define mpii_read_reply_free(s, v)   mpii_read((s), \
> +#define mpii_read_reply_free(s)              mpii_read((s), \
>                                               MPII_REPLY_FREE_HOST_INDEX)
>  #define mpii_write_reply_free(s, v)  mpii_write((s), \
>                                               MPII_REPLY_FREE_HOST_INDEX, (v))
> -#define mpii_read_reply_post(s, v)   mpii_read((s), \
> +#define mpii_read_reply_post(s)              mpii_read((s), \
>                                               MPII_REPLY_POST_HOST_INDEX)
>  #define mpii_write_reply_post(s, v)  mpii_write((s), \
>                                               MPII_REPLY_POST_HOST_INDEX, (v))
> @@ -2289,11 +2310,13 @@ mpii_attach(struct mpii_softc *sc)
>               goto free_queues;
>       }
>  
> +#if 0
>       if (mpii_get_dpm(sc) != 0) {
>               printf("%s: unable to get driver persistent mapping\n",
>                   DEVNAME(sc));
>               goto free_queues;
>       }
> +#endif
>  
>       if (mpii_cfg_coalescing(sc) != 0) {
>               printf("%s: unable to configure coalescing\n", DEVNAME(sc));
> @@ -2314,19 +2337,17 @@ mpii_attach(struct mpii_softc *sc)
>                   DEVNAME(sc));
>               goto free_dpm;
>       }
> -     
> -     /* enable interrupts */
> -     mpii_write(sc, MPII_INTR_MASK, MPII_INTR_MASK_DOORBELL 
> -         | MPII_INTR_MASK_RESET);
> -     
> +
>       if (mpii_portenable(sc) != 0) {
>               printf("%s: unable to enable port\n", DEVNAME(sc));
>               goto free_dev;
>       } /* assume all discovery events are complete by now */
>  
> +#if 0
>       if (sc->sc_discovery_in_progress)
>               printf("%s: warning: discovery still in progress\n", 
>                   DEVNAME(sc));
> +#endif
>  
>       if (mpii_get_bios_pg2(sc) != 0) {
>               printf("%s: unable to get bios page 2\n", DEVNAME(sc));
> @@ -2362,6 +2383,22 @@ mpii_attach(struct mpii_softc *sc)
>       sc->sc_scsibus = (struct scsibus_softc *) config_found(&sc->sc_dev,
>               &saa, scsiprint);
>  
> +     /* enable interrupts */
> +     mpii_write(sc, MPII_INTR_MASK, MPII_INTR_MASK_DOORBELL 
> +         | MPII_INTR_MASK_RESET);
> +
> +#if NBIO > 0
> +     if (bio_register(&sc->sc_dev, mpii_ioctl) != 0)
> +             panic("%s: controller registration failed", DEVNAME(sc));
> +     else
> +             sc->sc_ioctl = mpii_ioctl;
> +
> +#ifndef SMALL_KERNEL
> +     if (mpii_create_sensors(sc) != 0)
> +             printf("%s: unable to create sensors\n", DEVNAME(sc));
> +#endif
> +#endif
> +
>       return (0);
>  
>  free_bios_pg2:
> @@ -2408,23 +2445,49 @@ mpii_detach(struct mpii_softc *sc)
>  int
>  mpii_intr(void *arg)
>  {
> -     struct mpii_softc       *sc = arg;
> +     struct mpii_softc               *sc = arg;
> +     struct mpii_reply_descriptor    *rdp;
> +     u_int8_t                        reply_flags;
> +     int                             rv = 0;
>  
> -     if (mpii_reply(sc) < 0)
> -             return (0);
> +     do {
> +             bus_dmamap_sync(sc->sc_dmat, MPII_DMA_MAP(sc->sc_reply_postq),
> +                 0, 8 * sc->sc_reply_post_qdepth, BUS_DMASYNC_POSTWRITE);
>  
> -     while (mpii_reply(sc) >= 0)
> -             ;
> -     
> -     mpii_write_reply_post(sc, sc->sc_reply_post_host_index);
> +             rdp = &sc->sc_reply_postq_kva[sc->sc_reply_post_host_index];
> +             reply_flags = (u_int8_t)(rdp->reply_flags) & 
> +                 MPII_REPLY_DESCR_FLAGS_TYPE_MASK;
> +             if ((reply_flags == MPII_REPLY_DESCR_FLAGS_UNUSED))
> +                     break;
> +             if (rdp->type_dependent2 == 0xffffffff) {
> +                     /*
> +                      * ioc is still writing to the reply post queue
> +                      * race condition - bail!
> +                      */
> +                     printf("%s: ioc is writing a reply\n", DEVNAME(sc));
> +                     break;
> +             }
> +             mpii_reply(sc, rdp);
> +             sc->sc_reply_post_host_index =
> +                 (sc->sc_reply_post_host_index + 1) %
> +                 sc->sc_reply_post_qdepth;
> +             rv |= 1;
> +     } while (1);
>  
> -     return (1);
> +     if (rv)
> +             mpii_write_reply_post(sc, sc->sc_reply_post_host_index);
> +
> +     return (rv);
>  }
>  
>  void
>  mpii_timeout_xs(void *arg)
>  {
> -/* XXX */
> +     struct mpii_ccb         *ccb = arg;
> +     struct mpii_softc       *sc = ccb->ccb_sc;
> +
> +     printf("%s: xfer timeout, ccb %d state %d\n", DEVNAME(sc),
> +         ccb->ccb_smid, ccb->ccb_state);
>  }
>  
>  int
> @@ -2540,7 +2603,6 @@ mpii_load_xs(struct mpii_ccb *ccb)
>  void
>  mpii_minphys(struct buf *bp, struct scsi_link *sl)
>  {
> -     /* XXX */
>       minphys(bp);
>  }
>  
> @@ -2986,10 +3048,10 @@ mpii_iocfacts(struct mpii_softc *sc)
>        *         need to think this through as the specs
>        *         and other existing drivers contradict
>        */
> -     sc->sc_reply_post_qdepth = 128;
> -     sc->sc_request_depth = 128;
> -     sc->sc_num_reply_frames = 63;
> -     sc->sc_reply_free_qdepth = 64;
> +     sc->sc_reply_post_qdepth = 8192;
> +     sc->sc_request_depth = 1024;
> +     sc->sc_num_reply_frames = 1151;
> +     sc->sc_reply_free_qdepth = 1152;
>       
>       DNPRINTF(MPII_D_MISC, "%s: sc_request_depth: %d "
>           "sc_num_reply_frames: %d sc_reply_free_qdepth: %d "
> @@ -3063,7 +3125,7 @@ mpii_iocinit(struct mpii_softc *sc)
>  
>       hi_addr = (u_int32_t)
>           ((u_int64_t)MPII_DMA_DVA(sc->sc_replies) >> 32);
> -     iiq.system_reply_address_high = hi_addr;
> +     iiq.system_reply_address_high = htole32(hi_addr);
>  
>       iiq.system_request_frame_base_address = 
>           (u_int64_t)MPII_DMA_DVA(sc->sc_requests);
> @@ -3381,8 +3443,10 @@ mpii_eventnotify(struct mpii_softc *sc)
>       enq->event_masks[1] = htole32(0xfffffffc);
>       enq->event_masks[2] = htole32(0xffffffff);
>       enq->event_masks[3] = htole32(0xffffffff);
> -     
> +
> +     s = splbio();   
>       mpii_start(sc, ccb);
> +     splx(s);
>  
>       return (0);
>  }
> @@ -3418,9 +3482,6 @@ mpii_event_process_ir_cfg_change_list(st
>       ce = (struct mpii_evt_ir_cfg_element *)(ccl + 1);
>  
>       for (i = 0; i < ccl->num_elements; i++, ce++) {
> -
> -             for (volid = 0; volid < sc->sc_max_devices; volid++)
> -
>               type = (letoh16(ce->element_flags) & 
>                   MPII_EVT_IR_CFG_ELEMENT_EF_ELEMENT_TYPE_MASK);
>               
> @@ -3440,7 +3501,7 @@ mpii_event_process_ir_cfg_change_list(st
>                       
>                               if (sc->sc_mpii_dev[volid] == NULL) {
>                                       device = malloc(sizeof(struct 
> mpii_device), 
> -                                         M_DEVBUF, M_ZERO);
> +                                         M_DEVBUF, M_NOWAIT | M_ZERO);
>                                       if (device == NULL) {
>                                               printf("%s: 
> mpii_event_ir_cfg_change_list "
>                                                   "unable to allocate 
> mpii_device\n",
> @@ -3566,7 +3627,7 @@ mpii_event_process_sas_topo_change(struc
>                               }       
>  
>                       device = malloc(sizeof(struct mpii_device), 
> -                         M_DEVBUF, M_ZERO);
> +                         M_DEVBUF, M_NOWAIT | M_ZERO);
>                       if (device == NULL) {
>                               printf("%s: mpii_event_ir_cfg_change_list "
>                                   "unable to allocate mpii_device\n",
> @@ -3574,6 +3635,7 @@ mpii_event_process_sas_topo_change(struc
>                               break;
>                       }
>                       
> +                     sc->sc_pd_count++;              
>                       device->dev_handle = letoh16(pe->attached_dev_handle);
>                       device->phy_num = tcl->start_phy_num + i;
>                       device->type = MPII_DEV_TYPE_PD;
> @@ -3748,7 +3810,7 @@ mpii_get_bios_pg2(struct mpii_softc *sc)
>       }
>  
>       pagelen = hdr.page_length * 4;
> -     sc->sc_bios_pg2 = malloc(pagelen, M_TEMP, M_WAITOK|M_CANFAIL);
> +     sc->sc_bios_pg2 = malloc(pagelen, M_TEMP, M_NOWAIT | M_CANFAIL);
>       if (sc->sc_bios_pg2 == NULL) {
>               DNPRINTF(MPII_D_RAID, "%s: mpii_get_bios_pg2 unable to "
>                   "allocate space for BIOS page 2\n", DEVNAME(sc));
> @@ -3902,7 +3964,7 @@ mpii_get_dpm_pg0(struct mpii_softc *sc, 
>       pagelen = sizeof(struct mpii_ecfg_hdr) + sc->sc_max_dpm_entries * 
>           sizeof(struct mpii_dpm_entry);
>       
> -     dpm_page = malloc(pagelen, M_TEMP, M_WAITOK|M_CANFAIL);
> +     dpm_page = malloc(pagelen, M_TEMP, M_NOWAIT | M_CANFAIL);
>       if (dpm_page == NULL) {
>               DNPRINTF(MPII_D_MISC, "%s: mpii_get_dpm_pg0 unable to allocate "
>                   "space for device persistence mapping page 0\n", 
> DEVNAME(sc));
> @@ -4054,7 +4116,7 @@ mpii_get_ioc_pg8(struct mpii_softc *sc)
>  
>       pagelen = hdr.page_length * 4; /* dwords to bytes */
>       
> -     page = malloc(pagelen, M_TEMP, M_WAITOK|M_CANFAIL);
> +     page = malloc(pagelen, M_TEMP, M_NOWAIT | M_CANFAIL);
>       if (page == NULL) {
>               DNPRINTF(MPII_D_CFG, "%s: mpii_get_ioc_pg8 unable to allocate "
>                   "space for ioc config page 8\n", DEVNAME(sc));
> @@ -4079,10 +4141,12 @@ mpii_get_ioc_pg8(struct mpii_softc *sc)
>       DNPRINTF(MPII_D_CFG, "%s:  irvolumemappingflags: 0x%04x\n", 
>           DEVNAME(sc), letoh16(page->ir_volume_mapping_flags));
>  
> +#if 0
>       if (!(page->flags & MPII_IOC_PG8_FLAGS_ENCLOSURE_SLOT_MAPPING))
>               /* XXX we don't currently handle persistent mapping mode */
>               printf("%s: warning: controller requested device persistence "
>                   "mapping mode is not supported.\n");
> +#endif
>  
>       sc->sc_max_dpm_entries = page->max_persistent_entries;
>       sc->sc_dpm_enabled = (sc->sc_max_dpm_entries) ? 1 : 0;
> @@ -4183,8 +4247,12 @@ mpii_req_cfg_header(struct mpii_softc *s
>               splx(s);
>       }
>  
> -     if (ccb->ccb_rcb == NULL)
> -             panic("%s: unable to fetch config header\n", DEVNAME(sc));
> +     if (ccb->ccb_rcb == NULL) {
> +             s = splbio();
> +             mpii_put_ccb(sc, ccb);
> +             splx(s);
> +             return (1);
> +     }
>       cp = ccb->ccb_rcb->rcb_reply;
>  
>       DNPRINTF(MPII_D_MISC, "%s:  action: 0x%02x sgl_flags: 0x%02x "
> @@ -4219,8 +4287,10 @@ mpii_req_cfg_header(struct mpii_softc *s
>       } else
>               *hdr = cp->config_header;
>  
> +     s = splbio();
>       mpii_push_reply(sc, ccb->ccb_rcb->rcb_reply_dva);
>       mpii_put_ccb(sc, ccb);
> +     splx(s);
>  
>       return (rv);
>  }
> @@ -4277,7 +4347,7 @@ mpii_req_cfg_page(struct mpii_softc *sc,
>       cq->page_address = htole32(address);
>       cq->page_buffer.sg_hdr = htole32(MPII_SGE_FL_TYPE_SIMPLE |
>           MPII_SGE_FL_LAST | MPII_SGE_FL_EOB | MPII_SGE_FL_EOL |
> -         (page_length * 4) |
> +         MPII_SGE_FL_SIZE_64 | (page_length * 4) |
>           (read ? MPII_SGE_FL_DIR_IN : MPII_SGE_FL_DIR_OUT));
>  
>       /* bounce the page via the request space to avoid more bus_dma games */
> @@ -4309,7 +4379,9 @@ mpii_req_cfg_page(struct mpii_softc *sc,
>       }
>  
>       if (ccb->ccb_rcb == NULL) {
> +             s = splbio();
>               mpii_put_ccb(sc, ccb);
> +             splx(s);
>               return (1);
>       }
>       cp = ccb->ccb_rcb->rcb_reply;
> @@ -4339,99 +4411,64 @@ mpii_req_cfg_page(struct mpii_softc *sc,
>       else if (read)
>               bcopy(kva, page, len);
>  
> +     s = splbio();
>       mpii_push_reply(sc, ccb->ccb_rcb->rcb_reply_dva);
>       mpii_put_ccb(sc, ccb);
> +     splx(s);
>  
>       return (rv);
>  }
>  
>  int
> -mpii_reply(struct mpii_softc *sc)
> +mpii_reply(struct mpii_softc *sc, struct mpii_reply_descriptor *rdp)
>  {
> -     struct mpii_reply_descriptor    *rdp;
>       struct mpii_ccb         *ccb = NULL;
>       struct mpii_rcb         *rcb = NULL;
> -     struct mpii_msg_reply   *reply = NULL;
>       u_int8_t                reply_flags;
> -     u_int32_t               reply_dva, i;
> +     u_int32_t               i;
>       int                     smid;
>  
>  
>       DNPRINTF(MPII_D_INTR, "%s: mpii_reply\n", DEVNAME(sc));
>  
> -     /* XXX need to change to to be just the reply we expect to read */
> -     bus_dmamap_sync(sc->sc_dmat, MPII_DMA_MAP(sc->sc_reply_postq), 0, 
> -         8 * sc->sc_reply_post_qdepth, BUS_DMASYNC_POSTWRITE);
> -              
> -     rdp = &sc->sc_reply_postq_kva[sc->sc_reply_post_host_index];
> -
> -     reply_flags = (u_int8_t)(rdp->reply_flags) & 
> -         MPII_REPLY_DESCR_FLAGS_TYPE_MASK;
> -     
> -     if ((reply_flags == MPII_REPLY_DESCR_FLAGS_UNUSED))
> -             return (-1);
> -
> -     if (dwordn(rdp, 1) == 0xffffffff)
> -             /*
> -              * ioc is still writing to the reply post queue
> -              * race condition - bail!
> -              */
> -              return (-1);
> -              
>       DNPRINTF(MPII_D_INTR, "%s:  dword[0]: 0x%08x\n", DEVNAME(sc), 
>           letoh32(dwordn(rdp, 0)));
>       DNPRINTF(MPII_D_INTR, "%s:  dword[1]: 0x%08x\n", DEVNAME(sc), 
>           letoh32(dwordn(rdp, 1)));
>  
> -     switch (reply_flags) {
> -     case MPII_REPLY_DESCR_FLAGS_ADDRESS_REPLY:
> +     reply_flags = (u_int8_t)(rdp->reply_flags) & 
> +         MPII_REPLY_DESCR_FLAGS_TYPE_MASK;
> +
> +     /* smid */
> +     smid = letoh16(rdp->type_dependent1);
> +
> +     if (reply_flags == MPII_REPLY_DESCR_FLAGS_ADDRESS_REPLY) {
>               /* reply frame address */
> -             reply_dva = letoh32(rdp->type_dependent2);
> -             i = (reply_dva - (u_int32_t)MPII_DMA_DVA(sc->sc_replies)) /
> -                     MPII_REPLY_SIZE;
> -             
> +             i = (letoh32(rdp->type_dependent2) -
> +                 (u_int32_t)MPII_DMA_DVA(sc->sc_replies)) / MPII_REPLY_SIZE;
> +
>               bus_dmamap_sync(sc->sc_dmat,
>                   MPII_DMA_MAP(sc->sc_replies), MPII_REPLY_SIZE * i,
>                   MPII_REPLY_SIZE, BUS_DMASYNC_POSTREAD);
>               
>               rcb = &sc->sc_rcbs[i];
> -             reply = rcb->rcb_reply;
> -             /* fall through */
> -     default:
> -             /* smid */
> -              smid = letoh16(rdp->type_dependent1);
>       }
>  
>       DNPRINTF(MPII_D_INTR, "%s: mpii_reply reply_flags: %d smid: %d reply: 
> %p\n",
> -         DEVNAME(sc), reply_flags, smid, reply);
> +         DEVNAME(sc), reply_flags, smid, rcb->rcb_reply);
>  
> -     if (smid)  {
> -             ccb = &sc->sc_ccbs[smid - 1];
> -
> -             /* XXX why is this necessary ? */
> -             bus_dmamap_sync(sc->sc_dmat, MPII_DMA_MAP(sc->sc_requests),
> -                 ccb->ccb_offset, MPII_REQUEST_SIZE,
> -                 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
> -
> -             ccb->ccb_state = MPII_CCB_READY;
> -             ccb->ccb_rcb = rcb;
> -     }
> -
> -     DNPRINTF(MPII_D_INTR, "    rcb: 0x%04x\n", rcb);
> -
> -     dwordn(rdp, 0) = 0xffffffff;
> -     dwordn(rdp, 1) = 0xffffffff;
> +     memset(rdp, 0xff, sizeof(*rdp));
>  
>       bus_dmamap_sync(sc->sc_dmat, MPII_DMA_MAP(sc->sc_reply_postq), 
>           8 * sc->sc_reply_post_host_index, 8, 
>           BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
> -             
> -     sc->sc_reply_post_host_index = (sc->sc_reply_post_host_index + 1) %
> -         sc->sc_reply_post_qdepth;
>  
> -     if (smid)
> +     if (smid)  {
> +             ccb = &sc->sc_ccbs[smid - 1];
> +             ccb->ccb_state = MPII_CCB_READY;
> +             ccb->ccb_rcb = rcb;
>               ccb->ccb_done(ccb);
> -     else
> +     } else
>               mpii_event_process(sc, rcb->rcb_reply);
>  
>       return (smid);
> @@ -4465,11 +4502,11 @@ mpii_dmamem_alloc(struct mpii_softc *sc,
>           NULL, BUS_DMA_NOWAIT) != 0)
>               goto unmap;
>  
> -     bzero(mdm->mdm_kva, size);
> +     DNPRINTF(MPII_D_MEM, "  kva: %p  dva: %p  map: %p  size: %d\n",
> +         mdm->mdm_kva, mdm->mdm_map->dm_segs[0].ds_addr, mdm->mdm_map,
> +         size);
>  
> -     DNPRINTF(MPII_D_MEM, "%s: mpii_dmamem_alloc size: %d mdm: %#x "
> -         "map: %#x nsegs: %d segs: %#x kva: %x\n",
> -         DEVNAME(sc), size, mdm->mdm_map, nsegs, mdm->mdm_seg, mdm->mdm_kva);
> +     bzero(mdm->mdm_kva, size);
>  
>       return (mdm);
>  
> @@ -4506,7 +4543,7 @@ mpii_alloc_dev(struct mpii_softc *sc)
>  
>       sc->sc_mpii_dev = malloc(sc->sc_max_devices * 
>           sizeof(struct mpii_device *), M_DEVBUF, 
> -         M_WAITOK | M_CANFAIL | M_ZERO);
> +         M_NOWAIT | M_CANFAIL | M_ZERO);
>  
>       if (sc->sc_mpii_dev == NULL) 
>               return (1);
> @@ -4524,7 +4561,7 @@ mpii_alloc_ccbs(struct mpii_softc *sc)
>       TAILQ_INIT(&sc->sc_ccb_free);
>  
>       sc->sc_ccbs = malloc(sizeof(struct mpii_ccb) * (sc->sc_request_depth-1),
> -         M_DEVBUF, M_WAITOK | M_CANFAIL | M_ZERO);
> +         M_DEVBUF, M_NOWAIT | M_CANFAIL | M_ZERO);
>       if (sc->sc_ccbs == NULL) {
>               printf("%s: unable to allocate ccbs\n", DEVNAME(sc));
>               return (1);
> @@ -4593,6 +4630,7 @@ mpii_put_ccb(struct mpii_softc *sc, stru
>       ccb->ccb_state = MPII_CCB_FREE;
>       ccb->ccb_xs = NULL;
>       ccb->ccb_done = NULL;
> +     ccb->ccb_rcb = NULL;
>       bzero(ccb->ccb_cmd, MPII_REQUEST_SIZE);
>       TAILQ_INSERT_TAIL(&sc->sc_ccb_free, ccb, ccb_link);
>  }
> @@ -4623,7 +4661,7 @@ mpii_alloc_replies(struct mpii_softc *sc
>       DNPRINTF(MPII_D_MISC, "%s: mpii_alloc_replies\n", DEVNAME(sc));
>  
>       sc->sc_rcbs = malloc(sc->sc_num_reply_frames * sizeof(struct mpii_rcb),
> -         M_DEVBUF, M_WAITOK|M_CANFAIL);
> +         M_DEVBUF, M_NOWAIT | M_CANFAIL);
>       if (sc->sc_rcbs == NULL)
>               return (1);
>  
> @@ -4656,11 +4694,13 @@ mpii_push_replies(struct mpii_softc *sc)
>               mpii_push_reply(sc, rcb->rcb_reply_dva);
>       }
>  }
> +
>  void 
>  mpii_start(struct mpii_softc *sc, struct mpii_ccb *ccb)
>  {    
>       struct mpii_request_header      *rhp;
>       struct mpii_request_descriptor  descriptor;
> +     u_int32_t                       *rdp = (uint32_t *)&descriptor;
>  
>       DNPRINTF(MPII_D_RW, "%s: mpii_start %#x\n", DEVNAME(sc),
>           ccb->ccb_cmd_dva);
> @@ -4696,24 +4736,22 @@ mpii_start(struct mpii_softc *sc, struct
>       ccb->ccb_state = MPII_CCB_QUEUED;
>  
>       DNPRINTF(MPII_D_RW, "%s:   MPII_REQ_DESC_POST_LOW (0x%08x) write "
> -         "0x%08x\n", DEVNAME(sc), MPII_REQ_DESC_POST_LOW, 
> -         dwordn(&descriptor, 0));
> +         "0x%08x\n", DEVNAME(sc), MPII_REQ_DESC_POST_LOW, *rdp);
>  
>       DNPRINTF(MPII_D_RW, "%s:   MPII_REQ_DESC_POST_HIGH (0x%08x) write "
> -         "0x%08x\n", DEVNAME(sc), MPII_REQ_DESC_POST_HIGH, 
> -         dwordn(&descriptor, 1));
> +         "0x%08x\n", DEVNAME(sc), MPII_REQ_DESC_POST_HIGH, *(rdp+1)); 
>  
> -     /* XXX make this 64 bit? */
> -     mpii_write(sc, MPII_REQ_DESC_POST_LOW, htole32(dwordn(&descriptor, 0)));
> -     mpii_write(sc, MPII_REQ_DESC_POST_HIGH, 
> -         htole32(dwordn(&descriptor, 1)));
> +     mpii_write(sc, MPII_REQ_DESC_POST_LOW, htole32(*rdp));
> +     mpii_write(sc, MPII_REQ_DESC_POST_HIGH, htole32(*(rdp+1)));
>  }
>  
>  int
>  mpii_complete(struct mpii_softc *sc, struct mpii_ccb *ccb, int timeout)
>  {
> -     int                     smid = -1;
> -     
> +     struct mpii_reply_descriptor    *rdp;
> +     u_int8_t                        reply_flags;
> +     int                             smid = -1;
> +
>       DNPRINTF(MPII_D_INTR, "%s: mpii_complete timeout %d\n", DEVNAME(sc),
>           timeout); 
>       
> @@ -4726,19 +4764,35 @@ mpii_complete(struct mpii_softc *sc, str
>                       delay(1000);
>                       continue;
>               }
> -             
> -             smid = mpii_reply(sc);
> -     
> -             /* generates PCI write every completed reply, but
> -              * prevents deadlock waiting for specific smid
> -              */
> -             mpii_write_reply_post(sc, sc->sc_reply_post_host_index);
> -             
> -             DNPRINTF(MPII_D_INTR, "%s: mpii_complete call to mpii_reply 
> returned: %d\n",
> -                 DEVNAME(sc), smid);
>  
> +             bus_dmamap_sync(sc->sc_dmat, MPII_DMA_MAP(sc->sc_reply_postq),
> +                 0, 8 * sc->sc_reply_post_qdepth, BUS_DMASYNC_POSTWRITE);
> +
> +             rdp = &sc->sc_reply_postq_kva[sc->sc_reply_post_host_index];
> +             reply_flags = (u_int8_t)(rdp->reply_flags) & 
> +                 MPII_REPLY_DESCR_FLAGS_TYPE_MASK;
> +             if ((reply_flags == MPII_REPLY_DESCR_FLAGS_UNUSED))
> +                     continue;
> +             if (rdp->type_dependent2 == 0xffffffff) {
> +                     /*
> +                      * ioc is still writing to the reply post queue
> +                      * race condition - bail!
> +                      */
> +                     printf("%s: ioc is writing a reply\n", DEVNAME(sc));
> +                     continue;
> +             }
> +             smid = mpii_reply(sc, rdp);
> +
> +             DNPRINTF(MPII_D_INTR, "%s: mpii_complete call to mpii_reply"
> +                 "returned: %d\n", DEVNAME(sc), smid);
> +
> +             sc->sc_reply_post_host_index =
> +                 (sc->sc_reply_post_host_index + 1) %
> +                 sc->sc_reply_post_qdepth;
>       } while (ccb->ccb_smid != smid);
>  
> +     mpii_write_reply_post(sc, sc->sc_reply_post_host_index);
> +
>       return (0);
>  }
>  
> @@ -4851,27 +4905,6 @@ mpii_scsi_cmd(struct scsi_xfer *xs)
>       if (ccb == NULL)
>               return (NO_CCB);
>       
> -     /* XXX */
> -     if (sc->sc_mpii_dev[link->target] == NULL) {
> -             DNPRINTF(MPII_D_MAP, "%s: mpii_scsi_cmd nonexistent tid %d\n", 
> -                 DEVNAME(sc), link->target);
> -             return (99);
> -     }
> -
> -     /* XXX */
> -     if (sc->sc_mpii_dev[link->target]->flags & MPII_DEV_UNUSED) {
> -             DNPRINTF(MPII_D_MAP, "%s: mpii_scsi_cmd tid %d is "
> -                 "MPII_DEV_UNUSED\n", DEVNAME(sc), link->target);
> -             return (99);
> -     }
> -
> -     /* XXX */
> -     if (sc->sc_mpii_dev[link->target]->flags & MPII_DEV_HIDDEN) {
> -             DNPRINTF(MPII_D_MAP, "%s: mpii_scsi_cmd tid %d is "
> -                 "MPII_DEV_HIDDEN\n", DEVNAME(sc), link->target);
> -             return (99);
> -     }
> -
>       DNPRINTF(MPII_D_CMD, "%s: ccb_smid: %d xs->flags: 0x%x\n",
>           DEVNAME(sc), ccb->ccb_smid, xs->flags);
>  
> @@ -4885,7 +4918,8 @@ mpii_scsi_cmd(struct scsi_xfer *xs)
>       io->sense_buffer_length = sizeof(xs->sense);
>       io->sgl_offset0 = 24; /* XXX fix this */
>       io->io_flags = htole16(xs->cmdlen);
> -     io->dev_handle = htole16(sc->sc_mpii_dev[link->target]->dev_handle);
> +     ccb->ccb_dev_handle = io->dev_handle =
> +         htole16(sc->sc_mpii_dev[link->target]->dev_handle);
>       io->lun[0] = htobe16(link->lun);
>  
>       switch (xs->flags & (SCSI_DATA_IN | SCSI_DATA_OUT)) {
> @@ -4919,8 +4953,6 @@ mpii_scsi_cmd(struct scsi_xfer *xs)
>               return (COMPLETE);
>       }
>  
> -     timeout_set(&xs->stimeout, mpii_timeout_xs, ccb);
> -
>       DNPRINTF(MPII_D_CMD, "%s:  sizeof(mpii_msg_scsi_io): %d "
>           "sizeof(mpii_ccb_bundle): %d sge offset: 0x%02x\n",
>           DEVNAME(sc), sizeof(struct mpii_msg_scsi_io),
> @@ -4945,6 +4977,9 @@ mpii_scsi_cmd(struct scsi_xfer *xs)
>               return (COMPLETE);
>       }
>  
> +     timeout_set(&xs->stimeout, mpii_timeout_xs, ccb);
> +     timeout_add(&xs->stimeout, (xs->timeout * 1000) / hz);
> +
>       DNPRINTF(MPII_D_CMD, "%s:    mpii_scsi_cmd(): opcode: %02x datalen: %d "
>           "req_sense_len: %d\n", DEVNAME(sc), xs->cmd->opcode,
>           xs->datalen, xs->req_sense_length);
> @@ -4972,7 +5007,8 @@ mpii_scsi_cmd_done(struct mpii_ccb *ccb)
>               bus_dmamap_unload(sc->sc_dmat, dmap);
>       }
>  
> -     /* timeout_del */
> +     if (!(xs->flags & SCSI_POLL))
> +             timeout_del(&xs->stimeout);
>       xs->error = XS_NOERROR;
>       xs->resid = 0;
>       xs->flags |= ITSDONE;
> @@ -5044,6 +5080,8 @@ mpii_scsi_cmd_done(struct mpii_ccb *ccb)
>  
>       case MPII_IOCSTATUS_BUSY:
>       case MPII_IOCSTATUS_INSUFFICIENT_RESOURCES:
> +     case MPII_IOCSTATUS_SCSI_IOC_TERMINATED:
> +     case MPII_IOCSTATUS_SCSI_TASK_TERMINATED:
>               xs->error = XS_BUSY;
>               break;
>  
> @@ -5061,10 +5099,499 @@ mpii_scsi_cmd_done(struct mpii_ccb *ccb)
>               bcopy(&mcb->mcb_sense, &xs->sense, sizeof(xs->sense));
>  
>  
> -     DNPRINTF(MPII_D_CMD, "%s:  xs err: 0x%02x status: %d\n", DEVNAME(sc),
> +     DNPRINTF(MPII_D_CMD, "%s:  xs err: 0x%d status: %#x\n", DEVNAME(sc),
>           xs->error, xs->status);
>  
>       mpii_push_reply(sc, ccb->ccb_rcb->rcb_reply_dva);
>       mpii_put_ccb(sc, ccb);
>       scsi_done(xs);
>  }
> +
> +int
> +mpii_scsi_ioctl(struct scsi_link *link, u_long cmd, caddr_t addr, int flag,
> +    struct proc *p)
> +{
> +     struct mpii_softc       *sc = (struct mpii_softc *)link->adapter_softc;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_scsi_ioctl\n", DEVNAME(sc));
> +
> +     if (sc->sc_ioctl)
> +             return (sc->sc_ioctl(link->adapter_softc, cmd, addr));
> +     else
> +             return (ENOTTY);
> +}
> +
> +#if NBIO > 0
> +int
> +mpii_ioctl(struct device *dev, u_long cmd, caddr_t addr)
> +{
> +     struct mpii_softc       *sc = (struct mpii_softc *)dev;
> +     int                     error = 0;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_ioctl ", DEVNAME(sc));
> +
> +     switch (cmd) {
> +     case BIOCINQ:
> +             DNPRINTF(MPII_D_IOCTL, "inq\n");
> +             error = mpii_ioctl_inq(sc, (struct bioc_inq *)addr);
> +             break;
> +
> +     case BIOCVOL:
> +             DNPRINTF(MPII_D_IOCTL, "vol\n");
> +             error = mpii_ioctl_vol(sc, (struct bioc_vol *)addr);
> +             break;
> +
> +     case BIOCDISK:
> +             DNPRINTF(MPII_D_IOCTL, "disk\n");
> +             error = mpii_ioctl_disk(sc, (struct bioc_disk *)addr);
> +             break;
> +
> +     default:
> +             DNPRINTF(MPII_D_IOCTL, " invalid ioctl\n");
> +             error = EINVAL;
> +     }
> +
> +     return (error);
> +}
> +
> +int
> +mpii_ioctl_inq(struct mpii_softc *sc, struct bioc_inq *bi)
> +{
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_ioctl_inq\n", DEVNAME(sc));
> +
> +     strlcpy(bi->bi_dev, DEVNAME(sc), sizeof(bi->bi_dev));
> +
> +     bi->bi_nodisk = sc->sc_pd_count;
> +     bi->bi_novol = sc->sc_vd_count;
> +
> +     return (0);
> +}
> +
> +int
> +mpii_ioctl_vol(struct mpii_softc *sc, struct bioc_vol *bv)
> +{
> +     struct mpii_cfg_raid_vol_pg0    *vpg;
> +     struct mpii_cfg_hdr             hdr;
> +     struct scsi_link                *lnk;
> +     struct device                   *dev;
> +     size_t                          pagelen;
> +     u_int16_t                       volh;
> +     int                             rv, hcnt = 0;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_ioctl_vol %d\n",
> +         DEVNAME(sc), bv->bv_volid);
> +
> +     if (bv->bv_volid > sc->sc_vd_count)
> +             return (ENODEV);
> +
> +     volh = sc->sc_mpii_dev[sc->sc_vd_id_low + bv->bv_volid]->dev_handle;
> +
> +     if (mpii_req_cfg_header(sc, MPII_CONFIG_REQ_PAGE_TYPE_RAID_VOL, 0,
> +         MPII_CFG_RAID_VOL_ADDR_HANDLE | volh, 0, &hdr) != 0) {
> +             printf("%s: unable to fetch header for raid volume page 0\n",
> +                 DEVNAME(sc));
> +             return (EINVAL);
> +     }
> +
> +     pagelen = hdr.page_length * 4;
> +     vpg = malloc(pagelen, M_TEMP, M_WAITOK | M_CANFAIL | M_ZERO);
> +     if (vpg == NULL) {
> +             printf("%s: unable to allocate space for raid "
> +                 "volume page 0\n", DEVNAME(sc));
> +             return (ENOMEM);
> +     }
> +
> +     if (mpii_req_cfg_page(sc, MPII_CFG_RAID_VOL_ADDR_HANDLE | volh, 0,
> +         &hdr, 1, vpg, pagelen) != 0) {
> +             printf("%s: unable to fetch raid volume page 0\n",
> +                 DEVNAME(sc));
> +             free(vpg, M_TEMP);
> +             return (EINVAL);
> +     }
> +
> +     bv->bv_percent = -1;
> +
> +     switch (vpg->volume_state) {
> +     case MPII_CFG_RAID_VOL_0_STATE_ONLINE:
> +     case MPII_CFG_RAID_VOL_0_STATE_OPTIMAL:
> +             bv->bv_status = BIOC_SVONLINE;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_DEGRADED:
> +             bv->bv_status = BIOC_SVDEGRADED;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_FAILED:
> +             bv->bv_status = BIOC_SVOFFLINE;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_INITIALIZING:
> +             bv->bv_status = BIOC_SVBUILDING;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_MISSING:
> +             bv->bv_status = BIOC_SVINVALID;
> +             break;
> +     }
> +
> +     switch (vpg->volume_type) {
> +     case MPII_CFG_RAID_VOL_0_TYPE_RAID0:
> +             bv->bv_level = 0;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_TYPE_RAID1:
> +             bv->bv_level = 1;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_TYPE_RAID1E:
> +     case MPII_CFG_RAID_VOL_0_TYPE_RAID10:
> +             bv->bv_level = 10;
> +             break;
> +     default:
> +             bv->bv_level = -1;
> +             break;
> +     }
> +
> +     if ((rv = mpii_bio_hs(sc, NULL, 0, vpg->hot_spare_pool, &hcnt)) != 0) {
> +             free(vpg, M_TEMP);
> +             return (rv);
> +     }
> +
> +     bv->bv_nodisk = vpg->num_phys_disks + hcnt;
> +
> +     bv->bv_size = letoh64(vpg->max_lba) * letoh16(vpg->block_size);
> +
> +     lnk = sc->sc_scsibus->sc_link[bv->bv_volid][0];
> +     if (lnk != NULL) {
> +             dev = lnk->device_softc;
> +             strlcpy(bv->bv_dev, dev->dv_xname, sizeof(bv->bv_dev));
> +     }
> +
> +     free(vpg, M_TEMP);
> +     return (0);
> +}
> +
> +int
> +mpii_ioctl_disk(struct mpii_softc *sc, struct bioc_disk *bd)
> +{
> +     struct mpii_cfg_raid_vol_pg0            *vpg;
> +     struct mpii_cfg_raid_vol_pg0_physdisk   *pd;
> +     struct mpii_cfg_hdr                     hdr;
> +     size_t                                  pagelen;
> +     u_int16_t                               volh;
> +     u_int8_t                                dn;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_ioctl_disk %d/%d\n",
> +         DEVNAME(sc), bd->bd_volid, bd->bd_diskid);
> +
> +     if (bd->bd_volid > sc->sc_vd_count)
> +             return (ENODEV);
> +
> +     volh = sc->sc_mpii_dev[sc->sc_vd_id_low + bd->bd_volid]->dev_handle;
> +
> +     if (mpii_req_cfg_header(sc, MPII_CONFIG_REQ_PAGE_TYPE_RAID_VOL, 0,
> +         MPII_CFG_RAID_VOL_ADDR_HANDLE | volh, 0, &hdr) != 0) {
> +             printf("%s: unable to fetch header for raid volume page 0\n",
> +                 DEVNAME(sc));
> +             return (EINVAL);
> +     }
> +
> +     pagelen = hdr.page_length * 4;
> +     vpg = malloc(pagelen, M_TEMP, M_WAITOK | M_CANFAIL | M_ZERO);
> +     if (vpg == NULL) {
> +             printf("%s: unable to allocate space for raid "
> +                 "volume page 0\n", DEVNAME(sc));
> +             return (ENOMEM);
> +     }
> +
> +     if (mpii_req_cfg_page(sc, MPII_CFG_RAID_VOL_ADDR_HANDLE | volh, 0,
> +         &hdr, 1, vpg, pagelen) != 0) {
> +             printf("%s: unable to fetch raid volume page 0\n",
> +                 DEVNAME(sc));
> +             free(vpg, M_TEMP);
> +             return (EINVAL);
> +     }
> +
> +     if (bd->bd_diskid >= vpg->num_phys_disks) {
> +             u_int8_t        hsmap = vpg->hot_spare_pool;
> +
> +             free(vpg, M_TEMP);
> +             return (mpii_bio_hs(sc, bd, vpg->num_phys_disks, hsmap, NULL));
> +     }
> +
> +     pd = (struct mpii_cfg_raid_vol_pg0_physdisk *)(vpg + 1) +
> +         bd->bd_diskid;
> +     dn = pd->phys_disk_num;
> +
> +     free(vpg, M_TEMP);
> +     return (mpii_bio_disk(sc, bd, dn));
> +}
> +
> +int
> +mpii_bio_hs(struct mpii_softc *sc, struct bioc_disk *bd, int nvdsk,
> +     u_int8_t hsmap, int *hscnt)
> +{
> +     struct mpii_cfg_raid_config_pg0 *cpg;
> +     struct mpii_raid_config_element *el;
> +     struct mpii_ecfg_hdr            ehdr;
> +     size_t                          pagelen;
> +     int                             i, nhs = 0;
> +
> +     if (bd)
> +             DNPRINTF(MPII_D_IOCTL, "%s: mpii_bio_hs %d\n", DEVNAME(sc),
> +                 bd->bd_diskid - nvdsk);
> +     else
> +             DNPRINTF(MPII_D_IOCTL, "%s: mpii_bio_hs\n", DEVNAME(sc));
> +
> +     if (mpii_req_cfg_header(sc, MPII_CONFIG_REQ_PAGE_TYPE_RAID_CONFIG,
> +         0, MPII_CFG_RAID_CONFIG_ACTIVE_CONFIG, MPII_PG_EXTENDED,
> +         &ehdr) != 0) {
> +             printf("%s: unable to fetch header for raid config page 0\n",
> +                 DEVNAME(sc));
> +             return (EINVAL);
> +     }
> +
> +     pagelen = ehdr.ext_page_length * 4;
> +     cpg = malloc(pagelen, M_TEMP, M_WAITOK | M_CANFAIL | M_ZERO);
> +     if (cpg == NULL) {
> +             printf("%s: unable to allocate space for raid config page 0\n",
> +                 DEVNAME(sc));
> +             return (ENOMEM);
> +     }
> +
> +     if (mpii_req_cfg_page(sc, MPII_CFG_RAID_CONFIG_ACTIVE_CONFIG,
> +         MPII_PG_EXTENDED, &ehdr, 1, cpg, pagelen) != 0) {
> +             printf("%s: unable to fetch raid config page 0\n",
> +                 DEVNAME(sc));
> +             free(cpg, M_TEMP);
> +             return (ENXIO);
> +     }
> +
> +     el = (struct mpii_raid_config_element *)(cpg + 1);
> +     for (i = 0; i < cpg->num_elements; i++, el++) {
> +             if (ISSET(el->element_flags,
> +                 MPII_RAID_CONFIG_ELEMENT_FLAG_HSP_PHYS_DISK) &&
> +                 el->hot_spare_pool == hsmap) {
> +                     /*
> +                      * diskid comparison is based on the idea that all
> +                      * disks are counted by the bio(4) in sequence, thus
> +                      * substracting the number of disks in the volume
> +                      * from the diskid yields us a "relative" hotspare
> +                      * number, which is good enough for us.
> +                      */
> +                     if (bd != NULL && bd->bd_diskid == nhs + nvdsk) {
> +                             free(cpg, M_TEMP);
> +                             return (mpii_bio_disk(sc, bd,
> +                                 el->phys_disk_num));
> +                     }
> +                     nhs++;
> +             }
> +     }
> +
> +     if (hscnt)
> +             *hscnt = nhs;
> +
> +     free(cpg, M_TEMP);
> +     return (0);
> +}
> +
> +int
> +mpii_bio_disk(struct mpii_softc *sc, struct bioc_disk *bd, u_int8_t dn)
> +{
> +     struct mpii_cfg_raid_physdisk_pg0       ppg;
> +     struct mpii_cfg_hdr                     hdr;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_bio_disk %d\n",
> +         DEVNAME(sc), bd->bd_diskid);
> +
> +     if (bd->bd_volid > sc->sc_vd_count)
> +             return (ENODEV);
> +
> +     hdr.page_version = 0;
> +     hdr.page_length = sizeof(ppg) / 4;
> +     hdr.page_number = 0;
> +     hdr.page_type = MPII_CONFIG_REQ_PAGE_TYPE_RAID_PD;
> +
> +     if (mpii_req_cfg_page(sc, MPII_CFG_RAID_PHYS_DISK_ADDR_NUMBER | dn, 0,
> +         &hdr, 1, &ppg, sizeof(ppg)) != 0) {
> +             printf("%s: unable to fetch raid drive page 0\n",
> +                 DEVNAME(sc));
> +             return (EINVAL);
> +     }
> +
> +     bd->bd_channel = ppg.phys_disk_bus;
> +     bd->bd_target = ppg.phys_disk_num;
> +
> +     switch (ppg.phys_disk_state) {
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_ONLINE:
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_OPTIMAL:
> +             bd->bd_status = BIOC_SDONLINE;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_OFFLINE:
> +             bd->bd_status = BIOC_SDOFFLINE;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_DEGRADED:
> +             bd->bd_status = BIOC_SDFAILED;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_REBUILDING:
> +             bd->bd_status = BIOC_SDREBUILD;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_HOTSPARE:
> +             bd->bd_status = BIOC_SDHOTSPARE;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_NOTCONFIGURED:
> +             bd->bd_status = BIOC_SDUNUSED;
> +             break;
> +     case MPII_CFG_RAID_PHYDISK_0_STATE_NOTCOMPATIBLE:
> +             bd->bd_status = BIOC_SDINVALID;
> +             break;
> +     }
> +
> +     bd->bd_size = letoh64(ppg.dev_max_lba) * letoh16(ppg.block_size);
> +
> +     scsi_strvis(bd->bd_vendor, ppg.product_id, sizeof(ppg.product_id));
> +     scsi_strvis(bd->bd_serial, ppg.serial, sizeof(ppg.serial));
> +
> +     return (0);
> +}
> +
> +#ifndef SMALL_KERNEL
> +int
> +mpii_bio_volstate(struct mpii_softc *sc, struct bioc_vol *bv)
> +{
> +     struct mpii_cfg_raid_vol_pg0    *vpg;
> +     struct mpii_cfg_hdr             hdr;
> +     size_t                          pagelen;
> +     u_int16_t                       volh;
> +
> +     DNPRINTF(MPII_D_IOCTL, "%s: mpii_ioctl_vol %d\n",
> +         DEVNAME(sc), bv->bv_volid);
> +
> +     if (bv->bv_volid > sc->sc_vd_count)
> +             return (ENODEV);
> +
> +     volh = sc->sc_mpii_dev[sc->sc_vd_id_low + bv->bv_volid]->dev_handle;
> +
> +     if (mpii_cfg_header(sc, MPII_CONFIG_REQ_PAGE_TYPE_RAID_VOL, 0,
> +         MPII_CFG_RAID_VOL_ADDR_HANDLE | volh, &hdr) != 0) {
> +             printf("%s: unable to fetch header for raid volume page 0\n",
> +                 DEVNAME(sc));
> +             return (EINVAL);
> +     }
> +
> +     pagelen = hdr.page_length * 4;
> +     vpg = malloc(pagelen, M_TEMP, M_NOWAIT | M_CANFAIL | M_ZERO);
> +     if (vpg == NULL) {
> +             printf("%s: unable to allocate space for raid "
> +                 "volume page 0\n", DEVNAME(sc));
> +             return (ENOMEM);
> +     }
> +
> +     if (mpii_cfg_page(sc, MPII_CFG_RAID_VOL_ADDR_HANDLE | volh,
> +         &hdr, 1, vpg, pagelen) != 0) {
> +             printf("%s: unable to fetch raid volume page 0\n",
> +                 DEVNAME(sc));
> +             free(vpg, M_TEMP);
> +             return (EINVAL);
> +     }
> +
> +     switch (vpg->volume_state) {
> +     case MPII_CFG_RAID_VOL_0_STATE_ONLINE:
> +     case MPII_CFG_RAID_VOL_0_STATE_OPTIMAL:
> +             bv->bv_status = BIOC_SVONLINE;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_DEGRADED:
> +             bv->bv_status = BIOC_SVDEGRADED;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_FAILED:
> +             bv->bv_status = BIOC_SVOFFLINE;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_INITIALIZING:
> +             bv->bv_status = BIOC_SVBUILDING;
> +             break;
> +     case MPII_CFG_RAID_VOL_0_STATE_MISSING:
> +             bv->bv_status = BIOC_SVINVALID;
> +             break;
> +     }
> +
> +     free(vpg, M_TEMP);
> +     return (0);
> +}
> +
> +int
> +mpii_create_sensors(struct mpii_softc *sc)
> +{
> +     struct scsibus_softc    *ssc = sc->sc_scsibus;
> +     struct device           *dev;
> +     int                     i;
> +
> +     sc->sc_sensors = malloc(sizeof(struct ksensor) * sc->sc_vd_count,
> +         M_DEVBUF, M_WAITOK | M_ZERO);
> +     if (sc->sc_sensors == NULL)
> +             return (1);
> +
> +     strlcpy(sc->sc_sensordev.xname, DEVNAME(sc),
> +         sizeof(sc->sc_sensordev.xname));
> +
> +     for (i = 0; i < sc->sc_vd_count; i++) {
> +             if (ssc->sc_link[i][0] == NULL)
> +                     goto bad;
> +
> +             dev = ssc->sc_link[i][0]->device_softc;
> +
> +             sc->sc_sensors[i].type = SENSOR_DRIVE;
> +             sc->sc_sensors[i].status = SENSOR_S_UNKNOWN;
> +
> +             strlcpy(sc->sc_sensors[i].desc, dev->dv_xname,
> +                 sizeof(sc->sc_sensors[i].desc));
> +
> +             sensor_attach(&sc->sc_sensordev, &sc->sc_sensors[i]);
> +     }
> +
> +     if (sensor_task_register(sc, mpii_refresh_sensors, 10) == NULL)
> +             goto bad;
> +
> +     sensordev_install(&sc->sc_sensordev);
> +
> +     return (0);
> +
> +bad:
> +     free(sc->sc_sensors, M_DEVBUF);
> +
> +     return (1);
> +}
> +
> +void
> +mpii_refresh_sensors(void *arg)
> +{
> +     struct mpii_softc       *sc = arg;
> +     struct bioc_vol         bv;
> +     int                     i;
> +
> +     for (i = 0; i < sc->sc_vd_count; i++) {
> +             bzero(&bv, sizeof(bv));
> +             bv.bv_volid = i;
> +             if (mpii_bio_volstate(sc, &bv))
> +                     return;
> +
> +             switch(bv.bv_status) {
> +             case BIOC_SVOFFLINE:
> +                     sc->sc_sensors[i].value = SENSOR_DRIVE_FAIL;
> +                     sc->sc_sensors[i].status = SENSOR_S_CRIT;
> +                     break;
> +
> +             case BIOC_SVDEGRADED:
> +                     sc->sc_sensors[i].value = SENSOR_DRIVE_PFAIL;
> +                     sc->sc_sensors[i].status = SENSOR_S_WARN;
> +                     break;
> +
> +             case BIOC_SVSCRUB:
> +             case BIOC_SVONLINE:
> +                     sc->sc_sensors[i].value = SENSOR_DRIVE_ONLINE;
> +                     sc->sc_sensors[i].status = SENSOR_S_OK;
> +                     break;
> +
> +             case BIOC_SVINVALID:
> +                     /* FALLTRHOUGH */
> +             default:
> +                     sc->sc_sensors[i].value = 0; /* unknown */
> +                     sc->sc_sensors[i].status = SENSOR_S_UNKNOWN;
> +             }
> +     }
> +}
> +#endif /* SMALL_KERNEL */
> +#endif /* NBIO > 0 */

Reply via email to